Larger-System Submission Extension Runbook

Generated: 2026-05-20.

This runbook prepares the larger-system extension lanes requested for the submission paper stack and records the hardware jobs submitted from that preparation pass.

Budget Position

• Live IBM allocation remaining at probe time: 3160 s.
• Live IBM allocation after methods and FIM submissions: 2768 s.
• Live IBM allocation after methods, FIM, and Phase 3 submissions: 2652 s.
• Live IBM allocation after final post-submission probe: 2440 s.
• Conservative one-backend estimate for all n=6/n=8 QPU lanes: 1152 s.
• Conservative Fez+Marrakesh estimate for all n=6/n=8 QPU lanes: 2304 s.
• Estimated remaining margin after both backends: 856 s.
• Estimated full-stack margin after submitted methods/FIM lanes: 464 s.
• Estimated full-stack margin after submitted methods/FIM/Phase 3 lanes: 348 s.
• Estimated full-stack margin at final post-submission probe: 136 s.
• n=16 and n=20 are kept as local construction/transpilation scaling evidence, not QPU evidence.

The canonical readiness artifacts are:

• data/large_system_submission_extensions/large_system_submission_extension_readiness_20260520T193857Z.json
• data/large_system_submission_extensions/large_system_submission_extension_readiness_20260520T193903Z.json
• data/large_system_submission_extensions/large_system_submission_extension_readiness_20260520T193909Z.json
• docs/campaigns/large_system_submission_extension_readiness_20260520T193857Z.md
• docs/campaigns/large_system_submission_extension_readiness_20260520T193903Z.md
• docs/campaigns/large_system_submission_extension_readiness_20260520T193909Z.md

Prepared Candidate Order

1. phase3_reduced_pauli_n6
2. phase3_reduced_pauli_n8
3. fim_replication_zne_n6
4. fim_replication_zne_n8
5. methods_ansatz_energy_n6
6. methods_ansatz_energy_n8
7. methods_gpu_scaling_n16
8. methods_gpu_scaling_n20

Methods Paper Lanes

The methods-paper n=6 and n=8 IBM readiness lanes are prepared on Fez and Marrakesh. These are currently the direct submit-ready larger-system lanes.

Prepared readiness artifacts:

• data/rust_vqe_methods/ansatz_ibm_validation_readiness_ibm_fez_20260520T190928Z.json
• data/rust_vqe_methods/ansatz_ibm_validation_readiness_ibm_fez_20260520T190947Z.json
• data/rust_vqe_methods/ansatz_ibm_validation_readiness_ibm_marrakesh_20260520T191002Z.json
• data/rust_vqe_methods/ansatz_ibm_validation_readiness_ibm_marrakesh_20260520T191021Z.json

Submission commands, if the budget gate is still acceptable at execution time:

.venv-linux/bin/python scripts/submit_methods_ansatz_ibm_validation.py --backend ibm_fez --n-qubits 6 --physical-qubits 21 22 23 24 25 26 --max-qpu-seconds 3600 --experiment-id rust_vqe_methods_ansatz_validation_n6_fez_2026-05-20 --submit --confirm-budget
.venv-linux/bin/python scripts/submit_methods_ansatz_ibm_validation.py --backend ibm_fez --n-qubits 8 --physical-qubits 21 22 23 24 25 26 27 28 --max-qpu-seconds 3600 --experiment-id rust_vqe_methods_ansatz_validation_n8_fez_2026-05-20 --submit --confirm-budget
.venv-linux/bin/python scripts/submit_methods_ansatz_ibm_validation.py --backend ibm_marrakesh --n-qubits 6 --physical-qubits 2 3 4 5 6 7 --max-qpu-seconds 3600 --experiment-id rust_vqe_methods_ansatz_validation_n6_marrakesh_2026-05-20 --submit --confirm-budget
.venv-linux/bin/python scripts/submit_methods_ansatz_ibm_validation.py --backend ibm_marrakesh --n-qubits 8 --physical-qubits 1 2 3 4 5 6 7 8 --max-qpu-seconds 3600 --experiment-id rust_vqe_methods_ansatz_validation_n8_marrakesh_2026-05-20 --submit --confirm-budget

Submitted methods jobs:

Experiment	Backend	n	Job ID	Submission artifact
rust_vqe_methods_ansatz_validation_n6_fez_2026-05-20	ibm_fez	6	d870islg7okc73elnmmg	data/rust_vqe_methods/ansatz_ibm_validation_submission_ibm_fez_20260520T192017Z.json
rust_vqe_methods_ansatz_validation_n8_fez_2026-05-20	ibm_fez	8	d870j15g7okc73elnms0	data/rust_vqe_methods/ansatz_ibm_validation_submission_ibm_fez_20260520T192034Z.json
rust_vqe_methods_ansatz_validation_n6_marrakesh_2026-05-20	ibm_marrakesh	6	d870j52s46sc73f7j40g	data/rust_vqe_methods/ansatz_ibm_validation_submission_ibm_marrakesh_20260520T192050Z.json
rust_vqe_methods_ansatz_validation_n8_marrakesh_2026-05-20	ibm_marrakesh	8	d870j99789is73909fpg	data/rust_vqe_methods/ansatz_ibm_validation_submission_ibm_marrakesh_20260520T192107Z.json

FIM Hamiltonian Lanes

The FIM n=6 and n=8 Fez/Marrakesh replication/ZNE lanes were prepared and submitted after generalising scripts/submit_fim_replication_zne_ibm.py beyond the original n=4 execution path.

Submitted FIM jobs:

Experiment	Backend	n	Job ID	Submission artifact
scpn_fim_replication_zne_n6_fez_2026-05-20	ibm_fez	6	d870ld1789is73909hrg	data/scpn_fim_hamiltonian/fim_replication_zne_submission_ibm_fez_20260520T192537Z.json
scpn_fim_replication_zne_n8_fez_2026-05-20	ibm_fez	8	d870lhqs46sc73f7j6ag	data/scpn_fim_hamiltonian/fim_replication_zne_submission_ibm_fez_20260520T192555Z.json
scpn_fim_replication_zne_n6_marrakesh_2026-05-20	ibm_marrakesh	6	d870llp789is73909i60	data/scpn_fim_hamiltonian/fim_replication_zne_submission_ibm_marrakesh_20260520T192612Z.json
scpn_fim_replication_zne_n8_marrakesh_2026-05-20	ibm_marrakesh	8	d870lq9789is73909id0	data/scpn_fim_hamiltonian/fim_replication_zne_submission_ibm_marrakesh_20260520T192630Z.json

Methods n=16/n=20 Local Scaling

The n=16/n=20 scaling evidence was generated as local construction and transpilation evidence with error bars:

• data/rust_vqe_methods/vqe_convergence_methods_2026-05-20-large-system-extension.json
• data/rust_vqe_methods/vqe_convergence_aggregate_2026-05-20-large-system-extension.csv
• data/rust_vqe_methods/ansatz_scaling_error_bars_2026-05-20-large-system-extension.csv

This evidence must be described as local scaling only. It must not be described as IBM hardware evidence.

Phase 3 and FIM Larger-n Runner Status

The n=6/n=8 Phase 3 and FIM lanes are resource-ready by live Fez/Marrakesh transpilation probes. FIM is handled by the parameterised scripts/submit_fim_replication_zne_ibm.py; Phase 3 larger-system lanes are handled by dedicated scripts/submit_phase3_large_system_ibm.py.

Submitted Phase 3 larger-system jobs:

Experiment	Backend	n	Job ID	Submission artifact
phase3_large_system_n6_fez_2026-05-20	ibm_fez	6	d870qa8p0eas73dlj6kg	data/phase3_entanglement_tomography/phase3_large_system_submission_ibm_fez_20260520T193603Z.json
phase3_large_system_n8_fez_2026-05-20	ibm_fez	8	d870qf9789is73909nu0	data/phase3_entanglement_tomography/phase3_large_system_submission_ibm_fez_20260520T193622Z.json
phase3_large_system_n6_marrakesh_2026-05-20	ibm_marrakesh	6	d870qj9789is73909o2g	data/phase3_entanglement_tomography/phase3_large_system_submission_ibm_marrakesh_20260520T193640Z.json
phase3_large_system_n8_marrakesh_2026-05-20	ibm_marrakesh	8	d870qo0p0eas73dlj75g	data/phase3_entanglement_tomography/phase3_large_system_submission_ibm_marrakesh_20260520T193658Z.json

Retrieval and Reduction Status

All twelve submitted IBM jobs reported DONE in the bounded post-submission status probe and were retrieved.

Methods hardware ansatz validation

Backend	n	Raw-count artefact	Analysis artefact	Snapshot
ibm_fez	6	data/rust_vqe_methods/ansatz_ibm_validation_raw_counts_ibm_fez_20260520T195028Z.json	data/rust_vqe_methods/ansatz_ibm_validation_analysis_ibm_fez_20260520T195028Z.json	efficient_su2 has the lowest mitigated energy in this lane.
ibm_fez	8	data/rust_vqe_methods/ansatz_ibm_validation_raw_counts_ibm_fez_20260520T195040Z.json	data/rust_vqe_methods/ansatz_ibm_validation_analysis_ibm_fez_20260520T195040Z.json	efficient_su2 has the lowest mitigated energy in this lane.
ibm_marrakesh	6	data/rust_vqe_methods/ansatz_ibm_validation_raw_counts_ibm_marrakesh_20260520T195051Z.json	data/rust_vqe_methods/ansatz_ibm_validation_analysis_ibm_marrakesh_20260520T195051Z.json	efficient_su2 has the lowest mitigated energy in this lane.
ibm_marrakesh	8	data/rust_vqe_methods/ansatz_ibm_validation_raw_counts_ibm_marrakesh_20260520T195102Z.json	data/rust_vqe_methods/ansatz_ibm_validation_analysis_ibm_marrakesh_20260520T195102Z.json	efficient_su2 has the lowest mitigated energy in this lane.

This is valid n=6/n=8 hardware evidence for the methods paper, but it does not promote the topology-informed ansatz as backend-general superior on these larger lanes. The paper claim must say that the hardware validation was run and that the observed larger-lane ranking is ansatz- and backend-dependent.

FIM Hamiltonian n=6/n=8 replication/ZNE

Backend	n	Raw-count artefact	Analysis artefact	Mean abs raw scale-1 delta	Mean abs raw linear-ZNE delta
ibm_fez	6	data/scpn_fim_hamiltonian/fim_replication_zne_raw_counts_ibm_fez_20260520T195137Z.json	data/scpn_fim_hamiltonian/fim_replication_zne_analysis_ibm_fez_20260520T195137Z.json	0.4091	0.4021
ibm_fez	8	data/scpn_fim_hamiltonian/fim_replication_zne_raw_counts_ibm_fez_20260520T195149Z.json	data/scpn_fim_hamiltonian/fim_replication_zne_analysis_ibm_fez_20260520T195149Z.json	0.1479	0.1480
ibm_marrakesh	6	data/scpn_fim_hamiltonian/fim_replication_zne_raw_counts_ibm_marrakesh_20260520T195202Z.json	data/scpn_fim_hamiltonian/fim_replication_zne_analysis_ibm_marrakesh_20260520T195202Z.json	0.2474	0.2541
ibm_marrakesh	8	data/scpn_fim_hamiltonian/fim_replication_zne_raw_counts_ibm_marrakesh_20260520T195214Z.json	data/scpn_fim_hamiltonian/fim_replication_zne_analysis_ibm_marrakesh_20260520T195214Z.json	0.4231	0.4203

This closes the single-backend/no-ZNE criticism for the FIM paper without changing the claim boundary: FIM protection remains a falsified or backend-dependent hypothesis, not a promoted protection mechanism.

Phase 3 n=6/n=8 larger-system reduced-Pauli/ZNE lanes

Dedicated retrieval and dynamic-width reduction are implemented in scripts/retrieve_phase3_large_system_ibm.py.

Backend	n	Raw-count artefact	Analysis artefact	Linear-ZNE mean abs deviation	Readout-mitigated linear-ZNE mean abs deviation
ibm_fez	6	data/phase3_entanglement_tomography/phase3_large_system_raw_counts_20260520T195233Z_ibm_fez_n6.json	data/phase3_entanglement_tomography/phase3_large_system_analysis_20260520T195233Z_ibm_fez_n6.json	0.1688	0.1577
ibm_fez	8	data/phase3_entanglement_tomography/phase3_large_system_raw_counts_20260520T195245Z_ibm_fez_n8.json	data/phase3_entanglement_tomography/phase3_large_system_analysis_20260520T195245Z_ibm_fez_n8.json	0.1395	0.0980
ibm_marrakesh	6	data/phase3_entanglement_tomography/phase3_large_system_raw_counts_20260520T195300Z_ibm_marrakesh_n6.json	data/phase3_entanglement_tomography/phase3_large_system_analysis_20260520T195300Z_ibm_marrakesh_n6.json	0.3298	0.0688
ibm_marrakesh	8	data/phase3_entanglement_tomography/phase3_large_system_raw_counts_20260520T195312Z_ibm_marrakesh_n8.json	data/phase3_entanglement_tomography/phase3_large_system_analysis_20260520T195312Z_ibm_marrakesh_n8.json	0.3795	0.0604

Each Phase 3 lane has 36 reduced-Pauli channels. The n=6/n=8 data directly addresses the earlier "all hardware is n=4" limitation. Marrakesh shows a large readout-mitigation shift on the larger lanes, so the paper should treat this as a larger-width stress result with explicit readout sensitivity rather than as a simple monotone extension of the n=4 narrative.

Current Next Paper-Integration Slice

• Update the Phase 3 paper with the n=6/n=8 larger-system table, Kingston third-backend variant, Marrakesh n=8 repeat, and the readout-sensitivity boundary.
• Update the FIM paper with the second-backend, n=6/n=8, and ZNE replication table.
• Update the methods paper with the n=6/n=8 hardware validation result and the n=8 seed-23 sensitivity repeat while explicitly stating that the larger-lane ranking is not the same as the original n=4 topology-informed ansatz signal.
• Rebuild PDFs after the paper text and figure/table updates, not before.

Additional Variant Batch

The follow-up batch was executed after the first larger-system results to address whether one more useful variant class should be run while QPU time was still available. The added variants were:

Experiment	Backend	n	Job ID	Status	Analysis artefact
phase3_large_system_n6_kingston_2026-05-20	ibm_kingston	6	d871a3p789is7390a9p0	DONE	data/phase3_entanglement_tomography/phase3_large_system_analysis_20260520T201613Z_ibm_kingston_n6.json
phase3_large_system_n8_kingston_2026-05-20	ibm_kingston	8	d871ab8p0eas73dljo90	ERROR	Provider-side COS retrieval failure before quantum execution; quantum_seconds was zero.
phase3_large_system_n8_kingston_retry1_2026-05-20	ibm_kingston	8	d871bv1789is7390abng	DONE	data/phase3_entanglement_tomography/phase3_large_system_analysis_20260520T201817Z_ibm_kingston_n8.json
phase3_large_system_n8_marrakesh_readout_repeat_2026-05-20	ibm_marrakesh	8	d871aj1789is7390aa9g	DONE	data/phase3_entanglement_tomography/phase3_large_system_analysis_20260520T201657Z_ibm_marrakesh_n8.json
rust_vqe_methods_ansatz_validation_n8_seed23_fez_2026-05-20	ibm_fez	8	d871atas46sc73f7ju40	DONE	data/rust_vqe_methods/ansatz_ibm_validation_analysis_ibm_fez_20260520T201721Z.json
rust_vqe_methods_ansatz_validation_n8_seed23_marrakesh_2026-05-20	ibm_marrakesh	8	d871b90p0eas73dljpdg	DONE	data/rust_vqe_methods/ansatz_ibm_validation_analysis_ibm_marrakesh_20260520T201741Z.json

The Kingston Phase 3 lanes add a third-backend larger-width replication. The Kingston n=6 full readout calibration matrix was singular, so the reducer used a pseudoinverse replay and the paper treats that lane as a sensitivity boundary rather than ordinary full correlated readout mitigation. The Kingston n=8 retry completed with a pseudoinverse readout replay and a lower mitigated linear-ZNE mean deviation than the raw linear-ZNE estimate.

The Marrakesh n=8 repeat reproduced the earlier large raw/linear-ZNE deviation scale while again showing a small fully readout-mitigated linear-ZNE mean deviation. This strengthens the paper's readout-sensitivity boundary: the larger-width result is not a simple monotone extension of the n=4 reduced-Pauli story.

The methods seed-23 repeats preserve the larger-width ordering in which efficient_su2 has the lowest mitigated energy on both Fez and Marrakesh. This turns the methods hardware claim into a seed-sensitivity result rather than a single-seed anecdote.

After the batch, the live IBM usage probe reported 1577 remaining seconds for the account. No further QPU submissions were needed for this slice.